When Algorithms Replace Parties: AI Surveillance and Elite Power-Sharing in Autocracies

Authoritarian rulers have long struggled to govern amid uncertainty, unable to fully observe the loyalties and preferences of citizens and elites alike. Classic theories of the political economy of autocracy argue that dictators rely on parties, elections, and bureaucracies to overcome information asymmetries, using these institutions to monitor citizens, gauge elite loyalties, and sustain coalitions through power-sharing. This paper argues that AI surveillance technologies (such as facial recognition and predictive policing) fundamentally reconfigure this equilibrium. By enabling rulers to directly monitor social and elite behavior at unprecedented scale and precision, AI surveillance reduces the informational value of traditional institutions and weakens the need for elite consultation and power-sharing mechanisms. To test this argument, I draw on the AI & Big Data Global Surveillance Index by Steve Feldstein (compiling data from government and NGO reports, technology company records, and international media sources) to track the diffusion of AI surveillance technologies across 65 authoritarian regimes from the Global South and East between 2000 and 2023. I combine this data with indicators of party organization and elite cohesion from V-Party and estimate panel regressions with regime and year fixed effects. The results show that AI surveillance increases elite cohesion and personalism while eroding local party structures, membership, and societal linkages, reshaping the political economy of authoritarianism.